The yield curve is not a forecast
It is a price, set by people with positions, constraints and mandates. Reading it as the market's collective prediction attributes to it an intelligence it does not have.
The yield curve is probably the most over-interpreted object in finance. It gets described as the market's expectation of future rates, as a recession signal, as the aggregated intelligence of everyone with a view. It is treated, in short, as a forecast.
It is a price. The distinction matters more than it sounds.
Seven prices at seven tenors, each one clearing between buyers with entirely different objectives. Read as a forecast, that shape says rates rise steadily for thirty years, which nobody in the market actually believes. Read as a price, it says what money costs at each maturity today. Only the second reading is a question the curve can answer.
What a price is made of
A long-dated yield is not a clean statement about expected future short rates. It contains at least three things bundled together, and they are not separable by inspection.
There is an expectations component, the honest forecast piece. There is a term premium, the compensation demanded for holding duration, which moves with uncertainty and with the supply and demand for duration and has nothing to do with anyone's rate forecast. And there is a large block of flow that reflects neither: pension funds matching liabilities because regulation requires it, banks managing interest-rate risk on a book they did not choose, insurers with fixed mandates, index funds buying whatever the index holds, and central banks buying or selling for policy reasons that are explicitly not a view on value.
A substantial share of the buying in a government bond market is done by participants with no opinion about where rates are going. Their orders move the price identically to the orders of participants who do. What comes out the other end is a number, and the number is real, but calling it the market's expectation is attributing a view to people who do not have one.
| Component of a long yield | What moves it | Is it a forecast |
|---|---|---|
| Expectations | Views on the path of future short rates | Yes |
| Term premium | Uncertainty; the supply of and demand for duration | No |
| Mandated flow | Liability matching, index rules, ALM, policy purchases | No |
Half the buyers of a long bond are not expressing a view. They are complying with a mandate. The price does not distinguish between them.
The inversion case
The clearest illustration is the inversion-predicts-recession relationship, which is genuinely one of the better empirical regularities in macro and is routinely mis-stated.
The mechanism people describe, the market forecasting a downturn, is the weakest part of the story. The stronger mechanism is causal rather than predictive. An inverted curve compresses the margin on borrowing short and lending long, which is what a banking system does. Credit supply tightens. Tighter credit slows the economy. The curve is not observing the recession in advance so much as participating in creating it.
That reading changes what you should do with the signal. If the curve is a forecast, an inversion is information and you update. If it is part of the transmission mechanism, an inversion is a condition, and the question becomes how long it persists and how exposed your particular book is to the compression, which for a community bank funding long assets with short deposits is an operational question, not a macro one.
The distinction also explains why the relationship has been unreliable in periods when the term premium was being suppressed by large-scale official purchases. When the level of the long end is substantially set by a buyer with policy objectives, the shape carries less information about anything else. The signal did not stop working; the thing generating the signal changed.
A price is not a prediction
None of this makes the curve useless. It makes it a different object than the one it is usually treated as: a set of prices you can transact at, which is more useful than a forecast and answers a narrower question.
It says what funding costs at each tenor today, what a book earns and pays if nothing changes, and what a hedge costs. Those are exact answers. What it does not contain is a view, because most of the participants setting it do not have one.
The habit of reading a narrative out of the shape is hard to break, partly because the curve is one of the few market objects that looks like it is trying to tell you something. It is worth noticing how much of market commentary consists of attributing intent to an aggregation, and how reliably that attribution survives being wrong.
← Back to all insights